Ever spend several hours putting together what you think is a killer presentation, only to step back and realize that your visuals aren’t quite cutting it? That was me this weekend. There I was working on, what I thought was a killer project when my friend came in and said how my charts just really sucked. They were too complex and actually ended up taking away from the project instead of helping it. One of the trickier aspects with visualizations can be determining which visual technique best displays information. It’s important to know that each type of visualization has limits for how much it can display and still be useful – something I just learned the hard way.
After researching tips to create better visuals, I found this one article that ended up being super helpful – Best Practices: Maximum Elements For Different Visualization Types by Drew Skau. So, for those working on presentations or simply wanting to pick up some good rules of thumb for visuals, Skau outlines some excellent tips to help everyone out.
Pie Charts
According to Skau, “Pie Charts are among the most popular visualizations, but they aren’t appropriate for more than seven categories.” Just think about that for a minute, ever seen one of those pie charts with what seems like one hundred different slices – yep. They end up confusing the reader more than helping, “this is because our brains are not particularly good at telling different angles apart, and even worse at telling how far apart different angles are from each other.” Skau suggests that if you have a lot of different data that you don’t want to lump into an “other” category, that you may want to drop the pie chart and go with a bar chart instead.
Bar and Column Charts
Did you know that you can have too many bars in a bar chart? Maybe you did, but I didn’t. The best practices here aren’t as clear cut as with pie charts, but are equally beneficial. Skau suggests that when designing a bar chart you have to decide what matters most: the data, the overall trend, or the difference between the individual categories. Once you’ve decided what it is you are trying to communicate, then Skau suggests several best practices to help you out. “If the overall trend is the important factor, you might be able to get away with fifty bars of more. If the individual differences are important, you probably want to keep the total number of bars under twelve.” It’s important to remember that “Every bar you add increases the number of comparison possibilities exponentially” so making sure that you have thought out what it is you want to show off really plays a significant role.
Line Charts
Much like pie charts, having too many data sets graphed out in a line chart can make it difficult to read. Too many lines makes it hard for people to follow each line. Unlike pie or bar charts, there’s no rule of thumb to incorporate when designing line charts. (i.e. no rule of seven as is the case with pie charts and bar graphs) With line charts it really depends on what it is you are trying to communicate with the chart. For example, the chart below “only has seven lines, but the bottom ones cross a lot at low angles to each other…This isn’t a deal-breaker for this particular chart because the red line with the huge spike is where the story is, but not all data has the same story.”
Colors
You guessed it. Colors can be another limiting factor to visualizations. Similar to the overdone pie chart, having too many colors can make a graph difficult to read and therefore not pleasant. Skau suggests a best practice when using colors for pie chart wedges or bars in a bar graph is to use a maximum of twelve. This is, according to Skau, “The maximum number of colors (with similar luminance values) that we can distinguish and remember easily is around twelve.”
Skau points out that it’s important to keep in mind that “the more elements there are, the longer people will need to spend examining the visualization before they can interpret the information.” Ideally, a good rule of thumb when designing visualizations is the simpler. However, as Skau says, “the bottom line: if your data is too big, you may need more analysis to distill it down to the essential parts.”